Alvaro Herrera wrote:
> To be frank, I'm not sure I understand what you're saying here. I'm
> sure more analysis is good; that's easy to agree with.
>
> However, I don't want to be trapped in a design that's too hard to
> implement, or too hard for DBAs to manage.
+1
> There have been proposals to add these knobs:
> - maximum number of simultaneous processes (and make it more than 1)
> - times of day on which to change parameters (i.e. disable vacuum
> altogether, or make it more agressive or less)
Of all the autovacuum improvement discussions that I've seen recently
the one additional knob that sounds like a good idea to me is a way to
specify "HOT" tables that get special attention from autovacuum. There
has been a lot of conjecture as to a good implementation, and I don't
have a position on that, but I do think it's a good idea.